Advanced AI Demo Workflow for AI Developers Meetup

This n8n workflow demonstrates an advanced AI application, likely presented at an "AI Developers Meetup". It showcases a sophisticated integration of AI models, vector stores, and conversational memory to provide intelligent responses and perform actions based on user input. The workflow is designed to process chat messages, classify their intent, retrieve relevant information, and potentially interact with external APIs.

What it does

This workflow orchestrates several AI and automation components to deliver an intelligent conversational agent:

1. Listens for Chat Messages: It starts by receiving incoming chat messages, acting as the primary trigger for the workflow.
2. Classifies Message Intent: The incoming message is passed to a Text Classifier to determine its category or intent.
3. Conditional Routing based on Intent:
   • If the message is classified as "General Question", it proceeds to a Question and Answer Chain.
   • If the message is classified as "Tool Usage", it routes to an AI Agent designed to utilize tools.
   • If the message classification is neither of the above, it performs a "No Operation" (effectively ending the flow for unclassified messages).
4. Handles General Questions (Q&A Chain):
   • Utilizes a Pinecone Vector Store to retrieve relevant information.
   • Generates embeddings for the query using OpenAI Embeddings.
   • Processes the retrieved information and the user's question using an OpenAI Chat Model (or Anthropic Chat Model, if configured) to formulate an answer.
   • Maintains conversational context using a Simple Memory buffer.
5. Handles Tool Usage (AI Agent):
   • An AI Agent is employed to understand the user's request and decide if an external tool is needed.
   • It has access to an "HTTP Request Tool" which allows it to make API calls.
   • Maintains conversational context using a Simple Memory buffer.
   • Uses an OpenAI Chat Model (or Anthropic Chat Model, if configured) for reasoning and generating responses.
6. Prepares Documents for Vector Store (Initial Setup/Maintenance):
   • A separate branch, likely for initial data ingestion or updates, uses a "Default Data Loader" to load documents.
   • These documents are then split into manageable chunks using a "Recursive Character Text Splitter".
   • Embeddings are generated for these chunks using "Embeddings OpenAI".
   • Finally, these embeddings and document chunks are stored in the "Pinecone Vector Store".
7. Sends Email Notifications: In a separate, seemingly independent branch, the workflow can send emails via Gmail.
8. Sends Slack Messages: Another independent branch allows the workflow to post messages to Slack.
9. Performs HTTP Requests: A generic HTTP Request node is available, potentially for direct API calls outside of the AI Agent's tool usage.
10. Triggers on New Gmail Emails: The workflow can also be triggered by new emails received in a Gmail inbox.
11. Executes Custom Code: A Code node is included for executing custom JavaScript logic within the workflow.

Prerequisites/Requirements

To run this workflow, you will need:

• n8n Instance: A running instance of n8n.
• OpenAI API Key: For the OpenAI Chat Model and OpenAI Embeddings.
• Anthropic API Key (Optional): If you choose to use the Anthropic Chat Model.
• Pinecone Account and API Key: For the Pinecone Vector Store.
• Gmail Account: For the Gmail Trigger and Gmail node.
• Slack Account: For the Slack node.
• Credentials: Appropriate n8n credentials configured for OpenAI, Anthropic (optional), Pinecone, Gmail, and Slack.

Setup/Usage

1. Import the Workflow: Import the provided JSON into your n8n instance.
2. Configure Credentials:
   • Set up your OpenAI API Key credentials for the "OpenAI Chat Model" and "Embeddings OpenAI" nodes.
   • Set up your Anthropic API Key credentials for the "Anthropic Chat Model" node if you intend to use it.
   • Configure your Pinecone API Key and environment for the "Pinecone Vector Store" node.
   • Set up your Gmail OAuth 2.0 credentials for the "Gmail Trigger" and "Gmail" nodes.
   • Configure your Slack API Token credentials for the "Slack" node.
3. Activate the Workflow: Ensure the workflow is active to start listening for triggers.
4. Initial Document Ingestion (if applicable): If you are using the Pinecone Vector Store for RAG (Retrieval Augmented Generation), you'll need to run the "Default Data Loader" -> "Recursive Character Text Splitter" -> "Embeddings OpenAI" -> "Pinecone Vector Store" branch at least once to populate your vector database with relevant documents. This branch is currently not connected to the main chat flow, implying it's a separate setup step.
5. Trigger the Workflow:
   • Chat Trigger: Send a message to the configured chat platform that the "When chat message received" node is listening to.
   • Gmail Trigger: Receive a new email in the configured Gmail account.
   • Webhook: Make an HTTP POST request to the webhook URL provided by the "Webhook" node.
6. Monitor Execution: Observe the workflow execution in n8n to see the AI agent in action, classifying messages, retrieving information, and potentially interacting with external services.